CN-121981910-A - Unmanned aerial vehicle hyperspectral image stripe noise removing method and device, equipment and medium

CN121981910ACN 121981910 ACN121981910 ACN 121981910ACN-121981910-A

Abstract

The application provides a method, a device, equipment and a medium for removing hyperspectral image stripe noise of an unmanned aerial vehicle, which belong to the field of hyperspectral image stripe noise removal of unmanned aerial vehicles, and the method comprises the steps of obtaining a single-band gray image of a target area; the method comprises the steps of detecting edges of a gray image according to an edge detection operator to determine an external rectangle of the gray image, determining two sides and direction angles of each side which meet preset conditions according to the external rectangle, rotating the gray image according to the sides and the direction angles to obtain a rotated gray image, determining an effective area according to the rotated gray image, cutting the rotated gray image based on the minimum external rectangle of the effective area to obtain a cut gray image, and carrying out directional stripe noise filtering treatment on the cut gray image to obtain a filtered gray image. The application can accurately remove the stripe noise and completely retain the ground feature information.

Inventors

CUI HOUXIN
MA JUNJIE
WANG WEI
YU CAIHONG
DONG YANG

Assignees

河北先河环保科技股份有限公司
河北先进环保产业创新中心有限公司

Dates

Publication Date: 20260505
Application Date: 20260128

Claims (10)

1. The hyperspectral image stripe noise removing method for the unmanned aerial vehicle is characterized by comprising the following steps of: acquiring a hyperspectral image of a target area, and converting the hyperspectral image of the target area into a single-band gray level image; Detecting the edge of the gray image according to an edge detection operator, and determining the circumscribed rectangle of the gray image according to the edge of the gray image; determining two sides meeting preset conditions and direction angles of each side according to the circumscribed rectangle, and rotating the gray level image according to the sides and the direction angles to obtain a rotated gray level image, wherein the preset conditions are the first two sides arranged in descending order according to the length; Determining an effective area of the rotated gray image, and cutting the rotated gray image based on a minimum circumscribed rectangle of the effective area to obtain a cut gray image; and performing directional stripe noise filtering processing on the cut gray image to obtain a filtered gray image.
2. The method for removing hyperspectral image banding noise from an unmanned aerial vehicle according to claim 1, wherein detecting edges of the grayscale image according to an edge detection operator comprises: Performing edge detection on the gray level image according to the edge detection operator to obtain an edge pixel set; And extracting the edge of the gray scale image according to the edge pixel set.
3. The method for removing hyperspectral image banding noise of unmanned aerial vehicle according to claim 2, wherein the performing edge detection on the gray scale image according to the edge detection operator to obtain an edge pixel set comprises: determining gradient amplitude values among pixels in the gray scale image based on the gray scale image and the edge detection operator; taking the pixel with the gradient amplitude higher than a preset first gradient threshold value as a first edge pixel of the gray image; Taking pixels with gradient amplitude higher than a preset second gradient threshold and not higher than the preset first gradient threshold as second edge pixels of the gray scale image, wherein the preset first gradient threshold is higher than the preset second gradient threshold; and determining an edge pixel set of the gray scale image according to the first edge pixel and the second edge pixel.
4. The method for removing hyperspectral image banding noise of unmanned aerial vehicle according to claim 3, wherein determining the circumscribed rectangle of the gray scale image according to the edge of the gray scale image comprises: obtaining pixel points of each edge line segment according to the edges of the gray level images, and determining a pixel point data set according to each pixel point; connecting preset head and tail point coordinates in the pixel point data set to form a preset base line; executing vertex operation for determining an external rectangle based on the preset base line, and determining the external rectangle according to the vertex of the external rectangle; Wherein, the vertex operation for determining the circumscribed rectangle comprises: Obtaining the distance between each pixel point and the preset base line; if the maximum distance in each distance is higher than a preset precision threshold, taking the pixel point corresponding to the maximum distance as the vertex of the gray scale image; and if the distances are not higher than the preset precision threshold, taking the end point of the base line as the vertex of the gray scale image.
5. The method for removing hyperspectral image banding noise of unmanned aerial vehicle according to claim 1, wherein performing directional banding noise filtering processing on the clipped gray scale image to obtain a filtered gray scale image comprises: converting the cut gray scale image from a space domain to a frequency domain based on discrete Fourier transform to obtain frequency domain data of the gray scale image; Determining a mask for removing the vertical stripes according to the frequency domain data and a mask formula for removing the vertical stripes, wherein the mask formula for removing the vertical stripes is as follows: , wherein, A mask for removing vertical stripes, u for vertical frequency components, v for horizontal frequency components, M for the number of lines of the gray scale image, N for the number of columns of the gray scale image, Representing line width parameters, characterizing the bandwidth in the frequency domain that needs to be removed, Representing a center radius parameter; Determining a mask for removing the transverse stripes according to the frequency domain data and through a mask formula for removing the transverse stripes, wherein the mask formula for removing the transverse stripes is as follows: , wherein, A mask representing lateral stripe removal; determining a mask for removing stripe noise according to the mask for removing the vertical stripes and the mask for removing the horizontal stripes; and determining the gray level image after filtering according to the mask for removing the stripe noise.
6. The unmanned aerial vehicle hyperspectral image banding noise removal method of claim 5, further comprising, prior to said determining the filtered grayscale image from the banding noise removal mask: Carrying out centering treatment on the frequency domain data to obtain centered frequency domain data; the determining the filtered gray scale image according to the mask for removing the stripe noise comprises the following steps: Determining frequency domain data from which the stripe noise is removed according to the mask from which the stripe noise is removed and the frequency domain data after centering; and converting the frequency domain data from which the stripe noise is removed into a space domain based on inverse Fourier transform to obtain a filtered gray image.
7. The unmanned aerial vehicle hyperspectral image banding noise removal method of claim 1, wherein the method further comprises: Carrying out contrast stretching on the filtered gray image in a histogram matching mode to obtain a stretched gray image; And performing rotary reduction on the stretched gray scale image.
8. Unmanned aerial vehicle hyperspectral image stripe noise removing device, its characterized in that includes: the gray level image acquisition module is used for acquiring a hyperspectral image of a target area and converting the hyperspectral image of the target area into a single-band gray level image; the external rectangle determining module is used for detecting the edge of the gray image according to an edge detection operator and determining the external rectangle of the gray image according to the edge of the gray image; The image rotating module is used for determining two sides and direction angles of each side which meet preset conditions according to the circumscribed rectangle, and rotating the gray level image according to the sides and the direction angles to obtain a rotated gray level image, wherein the preset conditions are the first two sides which are arranged in descending order according to the length; the image clipping module is used for determining an effective area of the rotated gray image, clipping the rotated gray image based on the minimum circumscribed rectangle of the effective area and obtaining a clipped gray image; And the noise filtering module is used for carrying out directional stripe noise filtering processing on the cut gray image to obtain a filtered gray image.
9. An electronic device comprising a memory and a processor, the memory having a computer program stored therein, the processor, when running the computer program, performing the unmanned aerial vehicle hyperspectral image banding noise removal method of any one of claims 1 to 7.
10. A computer readable storage medium, characterized in that the computer readable storage medium has stored therein a computer program which, when executed by a processor, implements the unmanned aerial vehicle hyperspectral image banding noise removal method according to any one of claims 1 to 7.

Description

Unmanned aerial vehicle hyperspectral image stripe noise removing method and device, equipment and medium Technical Field The application belongs to the technical field of unmanned aerial vehicle hyperspectral image stripe noise removal, and particularly relates to a method, a device, equipment and a medium for removing hyperspectral image stripe noise of an unmanned aerial vehicle. Background The hyperspectral remote sensing technology can acquire continuous and fine spectral information of ground objects, and has irreplaceable advantages in the fields of quantitative analysis depending on spectral characteristics, such as river water quality monitoring, water ecological investigation, pollutant identification and the like. At present, an Unmanned aerial vehicle (Unmanned AERIAL VEHICLE, UAV) platform carrying a push-broom hyperspectral imager has become a main means for realizing high-resolution fine remote sensing of a river channel. In the push-broom imaging process, due to factors such as sensor probe response difference, circuit electronic interference, environmental change and the like, the generated stripe noise not only damages the visual uniformity and geometric texture of an image in space, but also seriously pollutes the authenticity of a ground object spectrum curve, so that systematic deviation is generated in quantitative ground object monitoring, and the application depth and reliability of hyperspectral data are greatly limited. At present, the method for removing the stripe noise mainly processes a single or synthesized wave band in a two-dimensional space domain, so that the calculation efficiency is improved, but the hyperspectral image has projection coordinate information, the stripe noise does not show a vertical state and a horizontal state, the conventional two-dimensional space domain processing method cannot fully consider the characteristic, and the inherent key characteristics of the stripe noise, which are in a linear structure in space and have high consistency among wave bands, are not fully utilized, so that the denoising process is not strong in pertinence. In addition, the method is easy to remove noise and simultaneously lose real ground object edge and texture information, or strip residues with directivity cannot be thoroughly eliminated, and the quality of hyperspectral images and the accuracy of subsequent analysis are affected. Disclosure of Invention The embodiment of the application aims to provide an unmanned aerial vehicle hyperspectral image stripe noise removing method, device, equipment and medium capable of accurately removing stripe noise and completely retaining ground object information. In order to achieve the above object, the technical solution provided by the embodiments of the present application is as follows: In a first aspect, a method for removing hyperspectral image stripe noise of an unmanned aerial vehicle is provided, including: Acquiring a hyperspectral image of a target area, and converting the hyperspectral image of the target area into a single-band gray level image; Detecting the edge of the gray image according to the edge detection operator, and determining the circumscribed rectangle of the gray image according to the edge of the gray image; Determining two sides and direction angles of each side which meet preset conditions according to the circumscribed rectangle, and rotating the gray level images according to the sides and the direction angles to obtain rotated gray level images, wherein the preset conditions are the first two sides which are arranged in descending order according to the length; Determining an effective area of the rotated gray image, and cutting the rotated gray image based on a minimum circumscribed rectangle of the effective area to obtain a cut gray image; And performing directional stripe noise filtering processing on the cut gray image to obtain a filtered gray image. In a second aspect, an apparatus for removing hyperspectral image stripe noise of an unmanned aerial vehicle is provided, including: The gray level image acquisition module is used for acquiring a hyperspectral image of the target area and converting the hyperspectral image of the target area into a single-band gray level image; The external rectangle determining module is used for detecting the edge of the gray image according to the edge detecting operator and determining the external rectangle of the gray image according to the edge of the gray image; The image rotating module is used for determining two sides and the direction angle of each side which meet the preset condition according to the external rectangle, and rotating the gray level image according to the sides and the direction angle to obtain a rotated gray level image, wherein the preset condition is the first two sides which are arranged in descending order according to the length; the image clipping module is used for determining an effective area of the rotated gray image, clip